Naza GPS yaw issue

Bison52

Member
Went out early this am in perfectly calm wind to test my GPS. Calm air is rare around here. GPS held a hover rock steady. Held position on climb and descent. Flew straight and then held on fore and aft flight and side to side.

But (ain't there always a but), on even a small gentle yaw, would drift off severely before correcting. Slow quarter turn to the left and it would drift 8 or ten feet to the left before slowing and returning to position. Same to the other direction. When it did correct, seemed to hunt for position a bit before returning to new original hover.

GPS lock is very solid. After I power on, by the time I step back and pick up tx, it has achieved 7 satellite lock.

Tried changing gain for yaw to no effect. May try recalibrating compass later

Anybody run into this yet?

Mike
 

Tahoe Ed

Active Member
I have not. An A/P friend of mine saw my video of PH/GPS yaw at max rotation. His comment was, that's impressive, but show me how it works with a really slow yaw. I am going to experiment with that. I am running 130 for my yaw if that helps.
 

Bison52

Member
Just wondering if there was any follow up. I went out today in a rare calm to very light wind day and flew some tests. In atti and GPS, hexa flew like it was on rails. Maybe a hint of the classic naza drift to the right but almost perfect atti hover. Perfectly steady on ascent to altitude in GPS and did well on descent.

But still has yaw issue in GPS mode. Make a quarter turn yaw and craft will slide off about 8 - 10 feet before correcting and returning to original position. I've adjusted gains and they are solid. Holds altitude well. Checked x,y,z entries for GPS and even moved GPS but no joy. Compass calibration is good too.

This is nothing I can't live with but it's so dialed in everywhere else that this kind of bugs me.

Oh yeah. Attempted a fail safe return to home from about 150 yards out and it landed right at my feet. So cool.

Mike
 

r4nd0m

Member
same here - I calibrated the compass and tried various settings - it seems to do it all the time. we have some gusty winds round here so it always needs to correct usually, I guess if it was 100% locked in the input would be too harsh ... I tested it today on our field and it even went for a circle with a diameter of nearly 20m at one time ...
 

Bison52

Member
What's odd is that it doesn't want to wander off until the yaw is completed. After the stick is moved to stop the yaw, it almost seems to hesitate for a moment, then wander off in pretty much a straight line. Then it slows and stops and heads back to where it belongs.

I thought at first it might be going in the direction of the wind, that is, it was losing position lock during the yaw and getting blown downwind before correcting but I don't think that's it. Need to do some more experimenting.
 


Have you guys tried rotating the GPS/Compass unit? See page 35 in the Naza manual. It relates to the problem you are describing. Try rotating the GPS unit opposite the direction you are seeing your MR fly.
 


Bison52

Member
Kevin. That is exactly what mine is doing. Let's figure this out. I've always used 10" props. DJI and Graupner.

What could be the difference between 8 and 10 props? Something about the torque difference? Rotating into prop wash?

My copter is in the bench while I'm making some upgrades. As soon as I'm back in the air I might try a few things. Wonder if increasing the throttle slightly to cause a slight rise with yaw has any affect?
 

OK so - some observations:

  • In my video, I yaw the quad clockwise in both cases (as seen from above) &, after the yawing is done, it drifts left/port (the lights are standard aircraft format: left/port = red, right/starboard = green. Front is white & the rear arms have the strobes)
  • In both cases, the yaw is approx. 180deg
  • In earlier tests, I have spun the quad on the spot & it doesnt drift until you stop yawing which is also your experience
It will be interesting to try 90/270/360/360+ & see what happens - I may try that tomorrow.

Other information:
  • On a quad, I assume yaw is achieved by slowing one pair of (same rotation) rotors & speeding the other (opposite rotation) pair - so overall lift remains constant
  • I also assume that lateral movement is achieved by slowing the rotor on one arm & speeding the rotor on the opposite arm, so that the quad tilts slightly but overall lift remains the same
  • One difference between 8" & 10" prop's is that the 10" will rotate slower for the same amount of lift (yes?) - assuming the pitch is the same
So the drift behaviour could be for many reasons. For example, if the flight controller takes too long to update the left (or right) ESC after yawing, then the quad would tilt & drift. Etc.

It would be interesting to see a DJI rep. pick up this thread & maybe provide some guidance about what we could test for them.

Any thoughts?
 

Bison52

Member
Kevin, I agree that everything works just like you said. As you stated, in tilted flight, overall lift has to increase because you have vectored part of the thrust to provide propulsion in a direction. I can't help but believe this is some self inflicted problem like gains. I know there are a number of NAZA GPS users who have experienced very stable yaw. Tahoe Ed was an early user for Naza GPS and he commented on how remarkable the yaw performance was.
 

VINHEX

Member
This is interesting as I've been looking into this on the Wookong-M I thought this was yaw settings, but haven't had chance to correct yet. Will see this weekend when get a little time to config.

Manual Say's

"The way of tuning the yaw gain is the same as the way of adjusting tail gyro. If you want a fast stick reaction speed, increase the gain, otherwise decrease the gain. However, the spin of the multi-rotor is produced by counter toque force, and the magnitude of which is limited, therefore, large gain will not produce tail vibration like a helicopter, But severe reaction at the start or stop motors, which will affect stabilisation of the other directions."

Hence why i'm looking into gains.. Will see :nevreness:
 
Last edited by a moderator:

Acmach5

Member
I have this tendency too. I would say somewhere in the middle. Not bad but noticeable. I liken it to a new pilot that turns nose in and takes a second to correct for the new orientation. I must admit I find it almost comforting to keep my piloting skills from going completely out the window with how amazingly easy it is to fly these things. That being said I like precise tuning as most of us here do and would love to see a solution.
 

Kevin, I agree that everything works just like you said. As you stated, in tilted flight, overall lift has to increase because you have vectored part of the thrust to provide propulsion in a direction. I can't help but believe this is some self inflicted problem like gains. I know there are a number of NAZA GPS users who have experienced very stable yaw. Tahoe Ed was an early user for Naza GPS and he commented on how remarkable the yaw performance was.

Just FYI - My gains have stayed as they are out of the box: 100% on everything.

It would be interesting to see if it has this behaviour (the initial drift after yaw) in Atti. (no GPS) mode & manual mode which would rule-out anything related to the GPS upgrade.
 

I did a more comprehensive test: http://www.youtube.com/watch?v=xsV6vf20TTw

It's not really conclusive, but in my case it seems that, no matter which way you yaw or how much, once you stop yawing, the quad drifts 45 degrees left and forward (i.e. in the direction of the front/port rotor arm).

If the weather improves (!) I might try experimenting in Atti. only mode (not GPS).
 

I have same problem in gps mode!!! In Atti mode is the same but i think was normal because wind move that, but in gps it's not acceptable i think!
 

Bison52

Member
This morning early was an absolute rarity in West Texas -- perfectly dead calm air. I took advantage to test some more and wasn't really surprised that the yaw issue was in atti mode also, so its not a GPS issue.

I also experienced an old Naza bugaboo -- the infamous Naza drift to the right. There was a lot of posting here at one time here about Naza platforms experiencing a slow steady drift to the right in hover. You don't notice it in the slightest breeze because you instinctively correct for it as you correct for a breeze.

I don't recall anybody ever coming up with a definitive cause or solution for Naza drift. I think it was generally regarded as one of these idiosyncrasies that you just learn to live with. I wouldn't be surprised if our yaw problems are related.

You might want to do a search for postings about Naza drift.

Mike

PS -- I chuckled recently about a posting talking about the performance of their craft in a strong wind -- gusting up there around 12 - 15 mph. We don't even notice here till it gets above 30.
 

... wasn't really surprised that the yaw issue was in atti mode also, so its not a GPS issue.

I was showing (& explaining) that video to a friend today & I noticed an interesting thing: when the quad drifts off, it drifts with a constant velocity, not a constant acceleration. So whatever causes it, is some kind of one-off impulse when the yaw stops, not a constant attitude offset. In my case, the drift always seems to be the same: i.e. in the direction of the forward/port motor arm.

So, one possible VERY SPECULATIVE cause could be this (btw: one of my past career activites was developing real-time system control software):
  • The FC has 3-axis gyro's - so it can constantly accumulate pitch/roll/yaw data &, in the absence of any control inputs, adjust the motors to keep that vector at zero
  • It can also estimate what vertical/lateral movement has occured as a result of those attitude changes & bring that back to zero too (although of course it can't know about external things like wind related drift - which is why we need the GPS to maintain accurate position lock)
  • When you yaw the quad (or generally), the FC does the above two things really well which is why it's such a good, stable platform
  • In an ideal world, assuming very sensitive gyros & the ability to truly, simultaneously monitor the inputs & control the motors (& no wind), the platform would never drift
  • However - in reality, the FC probably has a real-time operating system which time-slices the control functions
  • So - & here is the speculative part ... !
  • When you yaw the quad, one pair of motors speeds-up & the other pair slows down
  • During the yaw, the FC generally successfully keeps that cumulative attitude-vector-delta at zero (by successively receiving inputs from the gyros & sending correction changes to the motors in a time-sliced way)
  • When you stop yawing, the FC tries to revert the motors back to 'no yaw/fixed altitude' - but it does that in a time sliced way: motor-1, motor-2, motor-3 then motor-4
  • If there is an extra (or unconsidered) delay in reverting one of the motors, then that would manifest itself as a brief period of innappropriate attitude offset (until the motor-x controller actually sets the correct value) which would cause the quad to drift in a way which is undetectable by the FC alone
  • So the quad drifts at constant velocity until the GPS (which is on a much slower time constant) decides the position is awry & starts to pull it back
I'm sure there are many other possible causes but maybe whatever it is, is just too hard/not cost-effective to fix.
 

Bison52

Member
That explanation really sounds logical. My drift after yaw is not always in the same direction, depending on the degree and direction of the yaw but the difference may be that it's a hexa.

With just a layman's knowledge about the exact inner workings of the FC, I wonder if bad bearings or some other weakness in a motor could be causing it to "drag" slightly and cause the problem or maybe a weak connection in an ESC.
 

That's an interesting thought. Originally, one of my motors was more sluggish than the other three (it would stop very quickly) so I thought maybe that could be it. But I just checked & it no longer does that. Theoretically, it shouldn't make any difference - the FC should be able to track any outstanding vector offsets from sluggish motors & keep correcting until the attitude is back to zero & all consequential drift has been pulled back. Although sometimes, algorithms use predictive modelling to cut down the amount of incoming data needed & if the model didn't take account of things like a dragging motor that might cause it.

Anyway, it's not as if it's a huge problem! I still think the performance you get from the NAZA/GPS is amazing for the price. Compared with what was available 10 years ago!
 

Top